Studying is four jobs wearing one name
When you say you have been studying all afternoon, you could mean any of four completely different activities. You might have been hunting for material. You might have been sitting with something difficult, reading the same paragraph three times until it finally clicked. You might have been trying to get facts to stay in your head. Or you might have been producing something: an essay, a report, a set of answers, a presentation for Thursday.
Those four jobs feel like one job because they happen in the same session, at the same desk, with the same faintly anxious feeling. They are not one job. And AI is enormously helpful for some of them and quietly disastrous for others.
That split is the whole of this course. Everything else builds on it, so it is worth getting clear now.
In plain English
- Large language model:
- The thing behind most AI chat tools. It predicts likely text. It is not looking anything up unless it has been given a search tool.
- Hallucination:
- When a model states something false with complete confidence. Not a glitch, just what the system does when it has no better guess.
- Active recall:
- Testing yourself on something rather than rereading it. The single most useful study habit in this course.
- Academic integrity:
- Your institution's rules about what work must be your own. Every institution words these differently, so yours is the only version that counts.
Job one: finding material
You have a topic and you do not know where to start. Which search terms do specialists in this field actually use? What is the argument here, and who disagrees with whom? What is the name of the concept you have been circling for twenty minutes but cannot quite say?
This is a genuinely brilliant use. A model has absorbed an enormous amount of text about how subjects are organised, so it is very good at giving you vocabulary, sketching the shape of a debate, and turning your vague question into three precise ones. It saves you the hour you would otherwise spend working out what to type into the library catalogue.
There is one enormous trap sitting in the middle of this job, and it is important enough to be the whole of the next lesson: models will happily invent sources that do not exist. Use AI to generate search terms, not to hand you references.
Job two: understanding it
You have the material. You have opened the chapter or the paper and it may as well be in another language. Every sentence contains three words you would need to look up, and looking them up produces more words you would need to look up.
This is where AI is close to magical, and it is the use I would defend hardest. You can ask for an explanation pitched at exactly your level. You can ask for the same idea three different ways until one of them lands. You can ask what a specific sentence means, in context, at eleven at night when nobody is available. A patient explainer that never sighs, never makes you feel slow and never runs out of time is a real and serious advantage, particularly if you are studying alone or working in a second language.
It still gets things wrong, which is why Lesson 3 is half about getting good explanations and half about checking them.
Checkpoint
Studying is four separate jobs: finding material, understanding it, remembering it, and producing work. AI helps enormously with the first two.
Job three: remembering it
Understanding something in the moment and being able to retrieve it in three weeks are different skills, and the second one is what exams and jobs actually test.
AI is useful here, but in a narrow way: it is an excellent question generator. Feed it your notes, ask for practice questions, and you have removed the friction that stops most people testing themselves at all. That is a genuine gain, because the hard part of active recall was never the theory. It is that writing your own questions is tedious and you would rather reread the chapter.
What it cannot do is the remembering. You have to sit there and try to answer, badly, before you look. If you read the questions and the answers together, you have done reading, not recall, and you will feel far better prepared than you are.
Job four: producing work
Here is where it turns.
If the work is being marked, it exists to measure something about you. If a model writes it, the measurement is meaningless, and the person most affected is not your tutor. It is you. You are spending time, and usually money, in order to become someone who can do a particular thing. Getting a machine to do that thing produces a document and no person.
There is a plainer risk alongside that one. Submitting generated text as your own is a disciplinary matter at most institutions, the rules vary enormously between them, and detection is unreliable in ways that hurt honest students as well as dishonest ones. Lesson 5 deals with all of that properly, without the lecture.
The frame, in one table
| Job | What AI does for it | The risk if you lean too hard |
|---|---|---|
| Finding material | Excellent, for search terms and mapping a topic | Invented references you never opened |
| Understanding it | Excellent, at any level you ask for | A confident explanation that happens to be wrong |
| Remembering it | Useful, mainly as a question generator | Reading answers instead of recalling them |
| Producing work | Handle with care | You did not learn the thing, and you may be breaking the rules |
A quick test you can apply to any request before you send it: am I asking this to help me think, or asking it to think instead of me? The first is what this course teaches. The second is the thing your whole course of study exists to prevent.
The most underrated item on the list is asking for the same explanation three different ways. It costs you about thirty seconds and it works far more often than it has any right to.
Checkpoint
AI is a research and comprehension tool first, a revision tool second, and a writing tool only within whatever limits your institution sets.
๐ Quiz
Question 1 of 3Which of the four study jobs is AI most reliably useful for?